1. Field of the Invention
The present invention relates to a method for updating a filtering database in a multi-ring network having a hierarchical structure and, more particularly, to a technique that prevents the indiscriminative erasure of a filtering database by utilizing a control message for updating the filtering database when the topology of a lower ring network is altered in a multi-ring network in which a plurality of ring networks are connected by two ring connection nodes, thus allowing for effective, rapid protection switching.
2. Description of the Related Art
An Ethernet ring network prevents loop formation when traffic is transferred, by blocking a predetermined particular port among several ports constituting a ring. When the Ethernet ring network is faulty, the blocked port is open, while the port having the fault is blocked, altering network topology. In this case, in general, every node belonging to the ring then erases a filtering database in order to transfer a packet, according to the altered network topology.
A multi-Ethernet ring network in which a plurality of Ethernet ring networks are connected to each other has a structure such that a major ring is positioned at an upper side and a sub-ring is positioned at a lower side. Another lower sub-ring may be connected to the sub-ring. In other words, sub-rings may be connected to a single major ring, and further sub-rings may be connected to the sub-rings connected to the major ring to thereby extend the network.
In a multi-Ethernet ring network according to the related art, when the topology of a sub-ring is altered due to a link fault and recovery, the filtering tables of respective nodes belonging to the topology-altered sub-ring and the filtering tables of respective nodes belonging to a sub-ring or a major ring positioned at an upper side of the topology-altered sub-ring are all erased.
In this case, each node learns a source address of a frame so as to update the filtering database, and if a frame has a destination which is a new address not recorded in the filtering database, each node duplicates the corresponding frame and transfers it to all of the ports, excluding a port from which the frame has been received. Thus, in an Ethernet ring network according to the related art, a larger amount of frames whose address are not recoded are transmitted and received than those in a normal state until such time as the source address is learnt by all of the nodes. Namely, the related art multi-Ethernet ring network protection switching technique has a problem, in that a transition phenomenon occurs as the amount of traffic, after protection switching, is overshot.
Thus, in order to avoid frame loss caused by the transition phenomenon, a link capacity or a bandwidth of the network must be secured at a level greater than is strictly necessary, which is, however, ineffective. Meanwhile, when the link capacity or the bandwidth of the network are limited, a large capacity buffer may be provided so as to avoid such a frame loss, but in this case, rapid protection switching, e.g., protection switching within the 50 ms required for a real time voice communications network, cannot be provided. Therefore, a technique allowing for rapid protection switching in the event of an occurrence of a failure in the multi-Ethernet ring network having a limited link capacity or bandwidth is required.